基于孟德尔随机化方法探究多种危险因素与胰腺癌发病风险的因果关系
Exploring the Causal Relationship between Multiple Risk Factors and the Risk of Pancreatic Cancer Based on Mendelian Randomization Method

目的：胰腺癌是全球常见的死亡原因，可由多种危险因素引起。而这些因素相互关联、互为混杂。分离混杂因素得出因果关系，就可以了解是哪些因素对于胰腺癌的发生起到了实际影响。为此本研究利用两样本孟德尔随机化方法探究多种危险因素与胰腺癌发病风险之间的因果关系。方法：采用公共数据库中的工具变量(IV)分别进行两样本孟德尔随机化(TSMR)分析，通过逆方差加权法(IVW)，MR-Egger回归分析、加权中位数分析(WME)等两样本孟德尔随机化方法来评估肥胖、吸烟、饮酒、高血压、糖尿病及高胆固醇血症与胰腺癌发病风险之间的因果关系。最后进行敏感性分析来评估结果的可靠性与稳健性。结果：逆方差加权法显示肥胖(OR = 1.250, P = 0.04)、吸烟(OR = 5.386, P = 0.01)、饮酒(OR = 1.027, P = 0.39)、高血压(OR = 0.778, P = 0.004)、糖尿病(OR = 1.041, P = 0.51)、高胆固醇血症(OR = 0.480, P = 0.28)，数据显示吸烟、肥胖、高血压与胰腺癌的发病具有因果关系。敏感性分析的多效性及异质性P值均 > 0.05，因此孟德尔随机化在本研究中为因果推断的有效方法。结论：利用两样本孟德尔随机化方法排除混杂因素和反向因果关联，据此可以确定吸烟、肥胖、高血压与胰腺癌发病之间具有因果关系。
Objective: Pancreatic cancer is a common cause of mortality worldwide and can be influenced by various risk factors. These factors are interrelated and mutually affect each other. By separating the confounding factors and obtaining the causal relationship, we can know which factors play a real role in the occurrence of pancreatic cancer. This study aims to investigate the causal relationships between various risk factors and the risk of pancreatic cancer using two-sample Mendelian randomization (MR) methods. Methods: Two-sample Mendelian randomization (TSMR) analyses were conducted using instrumental variables (IV) from public databases. Causal relationships between obesity, smoking, alcohol consumption, hypertension, diabetes, and hypercholesterolemia and the risk of pancreatic cancer were evaluated using inverse variance-weighted (IVW), MR-Egger regression analysis, and weighted median analysis (WME) in the two-sample Mendelian randomization methods. Sensitivity analyses were performed to assess the reliability and robustness of the results. Results: Inverse variance-weighted analysis indicated that obesity (OR = 1.250, P = 0.04), smoking (OR = 5.386, P = 0.01), alcohol consumption (OR = 1.027, P = 0.39), hypertension (OR = 0.778, P = 0.004), diabetes (OR = 1.041, P = 0.51), and hypercholesterolemia (OR = 0.480, P = 0.28) showed causal relationships with the risk of pancreatic cancer. Sensitivity analyses demonstrated that the pleiotropy and heterogeneity P-values were both >0.05, indicating the effectiveness of Mendelian randomization for causal inference in this study. Conclusion: Two-sample Mendelian randomization methods, by eliminating confounding factors and reverse causality, confirm the causal relationships between